End-to-End n-ary Relation Extraction for Combination Drug Therapies

This is a code repository for our paper at IEEE ICHI 2023 titled: End-to-End n-ary Relation Extraction for Combination Drug Therapies. Please down the zipped file of this repo and follow the below instructions with the downloaded folder as the root directory.

Get access to the dataset

git clone https://github.com/allenai/drug-combo-extraction.git

Please see the paper A Dataset for N-ary Relation Extraction of Drug Combinations for more information about CombDrugExt dataset.

Preprocess dataset to our expected format

python preprocess.py

Installing Seq2rel using Poetry

curl -sSL https://install.python-poetry.org | python3 -
cd seq2rel
poetry install
cd ..

Training

Use the allennlp train command with .jsonnet config file we provid to train the model

train_data_path="drug-combo-extraction/n-ary/train.txt" \
valid_data_path="drug-combo-extraction/n-ary/valid.txt" \
dataset_size=1362 \
allennlp train "training-config-seq2rel/n-ary.jsonnet" \
    --serialization-dir "output"\
    --include-package "seq2rel" 

Evaluation

Example for evaluation on n-ary/test.txt file with fine-tuned model model.tar.gz with positive combination f1 score

python evaluation.py\
          --model output/model.tar.gz\
          --test_file drug-combo-extraction/n-ary/test.txt\
          --metric positive_combination_f1\